Traditional CI runs after you make changes. Iterate, validate locally, commit, push, wait.
With the rise of coding agents, this segmented workflow is too slow. It's not always feasible for an agent to validate its changes locally without the advantages of distributed, parallelized remote CI infrastructure. And nobody wants to review PRs produced by agents that hasn't produced a passing build yet.
RWX CLI v3.0.0 introduces new tools to tighten the feedback loop for developers and their agents:
- Start a run and wait for it to complete with
rwx run --wait - Inspect what happened with
rwx results - Grab what you need to troubleshoot failures with
rwx logsandrwx artifacts - All commands now support machine-readable
--output json
RWX's DAG-based execution and content-based caching already creates blazing fast feedback loops. And because you can trigger CI without the extra loop of pushing and pulling git commits or building container images, your changes begin validating immediately.
RWX tasks are inherently sandboxed and have explicitly specified dependencies. You can run an agent locally (or in the cloud) and have it validate its changes without it needing any local access to secrets or an up-to-date local development environment.
This means developers and agents can now iterate on a change until CI passes - without a single commit or push:
These changes are available starting with RWX CLI 3.0.0.
Related posts

Pulling task logs with the RWX CLI
The RWX CLI now supports pulling task logs with `rwx logs`, to power local debugging, automation, and AI assistants.

New Base Config to Run Directly from Container Images
We just shipped a change to how base images work on RWX. Base images are now configured directly from container images.

